Home // SECURWARE 2019, The Thirteenth International Conference on Emerging Security Information, Systems and Technologies // View article

Privacy Preserved Authentication: A Neural Network Approach

Authors:
Ray Hashemi
Amar Rasheed
Azita Bahrami
Jeffrey Young

Keywords: Anonymous Authentication; Encryption; Neural Network; Dynamic Authentication; Neural Network-based Authentication; Continuous Attesting Authenticity;

Abstract:
The anonymity of users during the authentication process for accessing computer-based Safety-Critical Systems (SCSs) are crucial for two reasons: (i) ever growing dependency of users on SCSs and (ii) Internet of Things, (IoT), social media, and marketers put the privacy of users of SCS in jeopardy more than ever. The goal of this research effort is to introduce and develop a novel neural network-based system that is able to (a) employ Extracted Eelectro-Cardiogram (ECG) feature vectors of the user as biometric credentials for authentication, (b) preserve privacy of users during the authentication process and (c) attest the authenticity of clients on a continuous basis during the time that SCS serves the client. Such attestation is necessary to make sure the user, after initial successful authentication, has not been replaced by an entity with malicious intent. Ten datasets with the total of 246,690 synthesized ECG feature vectors were created to test the system. These vectors were generated out of borrowed real ECG feature vectors for 90 users. Each dataset had 2,169 legitimate users’ credentials and 22,500 illegitimate ones. Our neural network-based system revealed the accuracy of (99.98%), precision of (100%), and sensitivity of (99.82%).

Pages: 16 to 21

Copyright: Copyright (c) IARIA, 2019

Publication date: October 27, 2019

Published in: conference

ISSN: 2162-2116

ISBN: 978-1-61208-746-7

Location: Nice, France

Dates: from October 27, 2019 to October 31, 2019